National Center for Computational Sciences

By: Scott Jones

Phoenix enables researchers to dissect dynamics of nature’s greatest aviator

Despite the progressive nature of science, many phenomena that we experience in our daily lives remain a mystery. Take hummingbirds for example.

Regardless of its prevalence in America’s backyards, little is understood about this creature’s ability to seemingly defy the laws of aerodynamics. These miniature marvels perform aerial feats that today’s fighter pilots can only dream of. By flapping their wings 15 to 80 times per second, depending on the species, hummingbirds can hover, fly backward (no other bird is capable of this), vertically, and laterally. Simply put, hummingbirds are excellent flyers—perhaps the best in the natural world.

"The mechanisms these animals employ to sustain flight do not seem to follow traditional aerodynamic theory and practice," said Andrew Johnson of Digital Rocket Science, who recently used the Phoenix supercomputer at Oak Ridge National Laboratory’s (ORNL’s) National Center for Computational Sciences (NCCS) to model and better understand the hummingbird’s flight. This research has the potential to greatly advance science in the areas of aerodynamics and biology. "These simulations and analysis could not take place without access to the ORNL/NCCS computers," said Johnson.

One potential application of this research is the more efficient design of micro air vehicles, tiny machines that could mimic hummingbirds to provide military and police units with improved surveillance, reconnaissance, and related activities. (Imagine a tiny hummingbird robot, manually controlled by some distant operator, clandestinely whizzing around a battlefield capturing images and video.) Johnson is comparing his hummingbird flight simulations with experimental data gathered from collaborators at Oregon State University (Doug Warrick) and the University of Portland (Bret Tobalske). However, because of the inherent limitations of experimental analysis, said Johnson, simulation is necessary to gain a more thorough understanding of the aerodynamics involved in the hummingbird’s flight.

Figure 1. Geometry of the computational hummingbird wing, dimensions, and location of the cross-sections shown in Figures 3 through 7. The wing itself is flat (i.e. has zero thickness) and this closely models a real hummingbird wing that has almost zero thickness. This figure doesn’t show it well, but the wing is strongly chambered. The hummingbird in this figure would be facing to the right.

For example, the collaborators at Oregon State use real hummingbirds in a laboratory wind tunnel, which produces plenty of useful data. However, said Johnson, you cannot tell the hummingbird what to do (in a laboratory) to study it at different angles and configurations; and because the team uses lasers to achieve a cross-sectional picture of air flow around the wings, the resulting images are sometimes limited in showing the full three-dimensional behavior of the flow. With simulation, he said, you can tell the "computational bird" to do just what you want it to, perfect for studying different phenomena, angles, and configurations.

"The results achieved from simulation are more comprehensive," said Johnson, adding that researchers can "plug in different wing types, frequencies, maneuvers, etc." He used approximately 10,000 hours on the NCCS’s Cray X1E Phoenix supercomputer to do just that. The end product: a series of movies that show air pressure on the top and bottom of the hummingbird wing, computed lift-and-drag curves, and animations of cross-section velocity vectors at various locations.

Figure 2. Lift and drag (side) forces for one of the hovering hummingbird simulations. The forces shown are for one wing. Three wing-beat cycles are shown. The higher lift bump is generated during the forward-stroke of the wing, and the smaller lift bump is generated during the back-stroke.

Phoenix is one of the few remaining global address-space vector systems in the world available for research. "The vector system at Oak Ridge National Laboratory is a great resource for the entire community," said Johnson. He further described his experience at the NCCS as "an ideal situation" because of Phoenix’s availability and an "efficient process." Furthermore, he said, all of the necessary documentation was readily available online, and all of his jobs got through quickly. When he did have a small problem, he said, the NCCS staff worked tirelessly to resolve it.

Figure 3. Velocity vectors at a cross-section 3.5cm from the center plane during the extent of the back-stroke of the wing. The colors and lengths of the vectors correspond to the fluid velocity.

Johnson’s results will be compared with the experimental results obtained by the Oregon team and with further experiments yet to be designed and carried out, possibly bringing researchers one step closer to technologically mimicking hummingbird flight and giving those in security situations a priceless advantage.